import pandas as pd
import numpy as np
import seaborn as sb
import matplotlib.pyplot as plt
import nltk
import os, re, csv
import inspect


import nltk.corpus
from nltk.corpus.reader.plaintext import PlaintextCorpusReader

from nltk.corpus import stopwords 
nltk.download('stopwords')
from nltk.tokenize import word_tokenize 
from collections import Counter

import nltk
nltk.download('punkt')

from nltk.stem import PorterStemmer 
from nltk.tokenize import word_tokenize 
nltk.download('vader_lexicon')
from nltk.sentiment.vader import SentimentIntensityAnalyzer
import matplotlib.cm as cm

from sklearn.cluster import MiniBatchKMeans
from sklearn.feature_extraction.text import TfidfVectorizer
from sklearn.decomposition import PCA
from sklearn.manifold import TSNE

%matplotlib inline

[nltk_data] Downloading package stopwords to
[nltk_data]     C:\Users\X\AppData\Roaming\nltk_data...
[nltk_data]   Package stopwords is already up-to-date!
[nltk_data] Downloading package punkt to
[nltk_data]     C:\Users\X\AppData\Roaming\nltk_data...
[nltk_data]   Package punkt is already up-to-date!
[nltk_data] Downloading package vader_lexicon to
[nltk_data]     C:\Users\X\AppData\Roaming\nltk_data...
[nltk_data]   Package vader_lexicon is already up-to-date!

merged_comments = pd.read_excel('./merged_comments.xlsx')
channel_topics = pd.read_excel('./yt_channels_topics.xlsx')

print(channel_topics)

              Channels                                        Assad  \
0          right-wing:                                          NaN   
1         Rubin Report                                          NaN   
2          Ben Shapiro  https://www.youtube.com/watch?v=lCh0Sti3Ons   
3       Steven Crowder  https://www.youtube.com/watch?v=EzUkoCRGNnc   
4   Paul Joseph Watson  https://www.youtube.com/watch?v=kmlrt_6MwXY   
5             The Hill                                          NaN   
6            centrist:                                          NaN   
7    Last Week Tonight                                          NaN   
8        Democracy Now  https://www.youtube.com/watch?v=SHHoCFyHzco   
9         David Pakman                                          NaN   
10         Young Turks  https://www.youtube.com/watch?v=89CpGuc73qc   
11          left-wing:                                          NaN   
12      Second Thought                                          NaN   
13               Vaush  https://www.youtube.com/watch?v=72lgT4ku4C4   
14     Majority Report  https://www.youtube.com/watch?v=d12ohmPAwzo   
15      F.D. Signifier                                          NaN   
16       news channels                                          NaN   
17                 CNN  https://www.youtube.com/watch?v=6R1JXjN05A8   
18                 BBC  https://www.youtube.com/watch?v=vtGswyr3OwQ   
19            Fox News  https://www.youtube.com/watch?v=9B_NOEe_4L0   

                              Health CEO murder  \
0                                           NaN   
1   https://www.youtube.com/watch?v=8XB2gjykF3M   
2   https://www.youtube.com/watch?v=Ecwxq9fDiQU   
3                                           NaN   
4                                           NaN   
5   https://www.youtube.com/watch?v=GSCSHMy52Jk   
6                                           NaN   
7                                           NaN   
8   https://www.youtube.com/watch?v=jyKJKUnMBY4   
9                                           NaN   
10  https://www.youtube.com/watch?v=Mlmq1rsNqg8   
11                                          NaN   
12                                          NaN   
13  https://www.youtube.com/watch?v=MhYWCXNXvbE   
14  https://www.youtube.com/watch?v=95hy5UE9sDk   
15                                          NaN   
16                                          NaN   
17  https://www.youtube.com/watch?v=fmmwF0hfi4I   
18  https://www.youtube.com/watch?v=ZioSHAcn6gM   
19  https://www.youtube.com/watch?v=nuFe2gOe_rc   

                Johnny Depp accused of violence  \
0                                           NaN   
1                                           NaN   
2   https://www.youtube.com/watch?v=bYpvVQnZvDU   
3   https://www.youtube.com/watch?v=vi6pvZcQ5AQ   
4                                           NaN   
5   https://www.youtube.com/watch?v=q-1KD2JWjdc   
6                                           NaN   
7                                           NaN   
8                                           NaN   
9   https://www.youtube.com/watch?v=GCjb-D-j2B8   
10  https://www.youtube.com/watch?v=p3vf67ChfN0   
11                                          NaN   
12                                          NaN   
13  https://www.youtube.com/watch?v=7SvYofqcdy4   
14  https://www.youtube.com/watch?v=ZuYVCDzhEAk   
15  https://www.youtube.com/watch?v=bblB5FtbnkU   
16                                          NaN   
17  https://www.youtube.com/watch?v=4LlO5sEpcF4   
18  https://www.youtube.com/watch?v=L8S-YWXGyv8   
19  https://www.youtube.com/watch?v=NWGTyS195og   

                                        Capitol  
0                                           NaN  
1   https://www.youtube.com/watch?v=QOXs-zT5v64  
2   https://www.youtube.com/watch?v=m8ZfmyxmgTA  
3   https://www.youtube.com/watch?v=6H_9ZY_Jabs  
4                                           NaN  
5                                           NaN  
6                                           NaN  
7   https://www.youtube.com/watch?v=0ri_Ma4DBks  
8   https://www.youtube.com/watch?v=F1sY2fQ5pnY  
9   https://www.youtube.com/watch?v=-NdutkXGc0U  
10  https://www.youtube.com/watch?v=gn4-LGIWMbE  
11                                          NaN  
12  https://www.youtube.com/watch?v=FoP9ufM3bjw  
13  https://www.youtube.com/watch?v=FoP9ufM3bjw  
14  https://www.youtube.com/watch?v=69opVkXYq2A  
15                                          NaN  
16                                          NaN  
17  https://www.youtube.com/watch?v=JLBLgH3PAeI  
18  https://www.youtube.com/watch?v=UXR_bqyAy4E  
19  https://www.youtube.com/watch?v=tVPSYr-xG6s

merged_comments.head()

merged_comments = merged_comments.drop(['User', 'Published At', 'Time Ago'], axis=1)
print(merged_comments.head())

                                             Comment  Likes channel  topic  \
0  BBC journalists must go to the doors of Britis...      0     bbc  assad   
1  Getting out of hands. It may sound off topic b...      0     bbc  assad   
2                                Imperialist go home      0     bbc  assad   
3  It's like we've seen this movie in the Middle ...      0     bbc  assad   
4  😂😂😂 how dumb and foolish the Americans and the...      0     bbc  assad   

  leaning  
0    news  
1    news  
2    news  
3    news  
4    news

merged_comments = merged_comments.dropna()

print(len(merged_comments))

113266

#unique values in the column as a vector or whatever for encoders later
uniqs = merged_comments['topic'].unique()
uniqs

array(['assad', 'ceo', 'depp', 'capitol'], dtype=object)

def splitter(channel: pd.DataFrame) -> pd.DataFrame:
    channel_comments = merged_comments.loc[merged_comments['topic'] == f'{channel}']
    channel_comments = channel_comments.drop(['topic'], axis=1)
    return channel_comments

list_of_names = {}

assad_comments = splitter('assad')
ceo_comments = splitter('ceo')
capitol_comments = splitter('capitol')
depp_comments = splitter('depp')
print(assad_comments.head())
print(assad_comments.dtypes)
print(depp_comments.head())

                                             Comment  Likes channel leaning
0  BBC journalists must go to the doors of Britis...      0     bbc    news
1  Getting out of hands. It may sound off topic b...      0     bbc    news
2                                Imperialist go home      0     bbc    news
3  It's like we've seen this movie in the Middle ...      0     bbc    news
4  😂😂😂 how dumb and foolish the Americans and the...      0     bbc    news
Comment    object
Likes       int64
channel    object
leaning    object
dtype: object
                                                Comment  Likes channel leaning
5268  Absolutely disgusting so called "journalism". ...      0     cnn    news
5269  Anyone who respects left-wing media should wat...      0     cnn    news
5270  CNN reporting is biased. Salute to USA justice...      0     cnn    news
5271  Amber Tu*d,was a se* worker,had affa*rs, with ...      0     cnn    news
5272  Believe all women, every day, all day and twic...      0     cnn    news

# replace special characters from text. 
# substituting "/", "@" and "|" and others by a space.

list_of_names = [assad_comments, depp_comments, capitol_comments, ceo_comments]

def cleaner(comments_clean: list):
    for x in comments_clean: 
        x = x.replace('[^a-zA-Z0-9 \n\.]', '', regex=True)
        #x = x.str.lower()

cleaner(list_of_names)
assad_comments['Comment'] = assad_comments['Comment'].str.lower()
ceo_comments['Comment'] = ceo_comments['Comment'].str.lower()
capitol_comments['Comment'] = capitol_comments['Comment'].str.lower()
depp_comments['Comment'] = depp_comments['Comment'].str.lower()

print(ceo_comments.head())

                                                Comment  Likes channel leaning
3237  well he was stressed out so of course he went ...      0     cnn    news
3238  police were  being  forceful  with him .  luig...      0     cnn    news
3239                       nullify nullify \nfree luigi      0     cnn    news
3240                         i am sad!! wish him well!!      0     cnn    news
3241                            hope we can save luigi.      0     cnn    news

ceo_comments['Likes']\
.unique()

array([   0,    2,    1,    3,    9,    4,    8,    5,   73,   22,   15,
         18,   10,  130,   53,  137,   14,   36,  153,   16,   19,    6,
         24,  281,    7,   43,  155,   12,   28,  198,   21,   42,  320,
         72,   31,  535,  224,   68,   25,  141,   62,   11,  181,   30,
         26,  215,  270,   63,  128,  790,   35,   87,   27,  227,  147,
        196,   52, 1184,   66,   34,  355,   47,   69,  203,  616,   20,
        622,  100,  856,  326,  286,  225,   45,  161,   29,  273,  347,
         67,  156,  105,  195,   38,  104,   48,  604,  127,   50,   13,
       2308,  117,  138,  480, 2260,  318,   39,  970,   76,  316, 3191,
       1898,  264,  660,  851,  385,  167,  282,  525,  748,  556,   37,
        208,   77,   23,   92,   44,   80,   65,  184,  640,  157, 1847,
        378,   32,  268,  499,  179, 1013,  862, 2695,  103,  135,  345,
         54,  114,  115,  375,  432, 1985,   51, 1000, 3507, 1949,   91,
         17,  457,   79,  794,  182, 1117,  221, 1850,   96, 4073,   82,
         46, 1012, 5671, 6776,  168,  232, 2844,  504,   86,  671,  106,
       4531,   40,  887,  491,  610,  160, 3169, 2128,  680,  315, 2990,
        251,   58,  172,  354,  708,  163,  124,  166,  183, 1314, 2061,
         56,   41,  112,  289, 1823,   98,  756,   59,  330,  263,  121,
        170,  154,   88,  236,  416,  442, 1157,   55,  219,  186,  430,
        536,  512,  911,  538,  995,  691,  209,  341,   99,  444,  193,
        142,  242,   61,  880,  204,  258, 4077, 1283,  864, 1833,  335,
        544,  279,   33,  148,  211,   85,  171, 1466,  144,  905,   60,
        713,  169,  420,  394,  107,   74,   71,  238,  146,  361,  125,
        132,  377,   70,   49,  417,   90,  628,  415,  118,   64,  126,
        119,   75,  129,  212,  331,  253, 1003,  244,   97,  323,  434,
        108,  590,  145,  191, 1407,  222, 1215,  207,  402,  687, 1121,
        541,  476,  406,  197,  311,   89,  486,  110,  228,  349,  175,
       1275,  470,  436,  654, 1245,  152,  809,  123,  245, 1839,  159,
        358,  408,   81,  220,  122,  241,  120,   93], dtype=int64)

tfidf = TfidfVectorizer(
    min_df = 5,
    max_df = 0.95,
    max_features = 8000,
    stop_words = 'english'
)
tfidf.fit(assad_comments['Comment'])
text_Assad = tfidf.transform(assad_comments['Comment'])
text_ceo = tfidf.transform(ceo_comments['Comment'])
text_capitol = tfidf.transform(capitol_comments['Comment'])
text_depp = tfidf.transform(depp_comments['Comment'])

def find_optimal_clusters(data, max_k):
    iters = range(2, max_k+1, 2)
    
    sse = []
    for k in iters:
        sse.append(MiniBatchKMeans(n_clusters=k, init_size=1024, batch_size=2048, random_state=20).fit(data).inertia_)
        print('Fit {} clusters'.format(k))
        
    f, ax = plt.subplots(1, 1)
    ax.plot(iters, sse, marker='o')
    ax.set_xlabel('Cluster Centers')
    ax.set_xticks(iters)
    ax.set_xticklabels(iters)
    ax.set_ylabel('SSE')
    ax.set_title(f'SSE by Cluster Center Plot of set')
    
find_optimal_clusters(text_Assad, 20)
find_optimal_clusters(text_ceo, 20)
find_optimal_clusters(text_capitol, 20)
find_optimal_clusters(text_depp, 20)

C:\Users\X\AppData\Roaming\Python\Python311\site-packages\sklearn\cluster\_kmeans.py:1934: FutureWarning: The default value of `n_init` will change from 3 to 'auto' in 1.4. Set the value of `n_init` explicitly to suppress the warning
  super()._check_params_vs_input(X, default_n_init=3)

Fit 2 clusters
Fit 4 clusters
Fit 6 clusters
Fit 8 clusters
Fit 10 clusters
Fit 12 clusters
Fit 14 clusters
Fit 16 clusters
Fit 18 clusters
Fit 20 clusters
Fit 2 clusters
Fit 4 clusters
Fit 6 clusters
Fit 8 clusters
Fit 10 clusters
Fit 12 clusters
Fit 14 clusters
Fit 16 clusters
Fit 18 clusters
Fit 20 clusters
Fit 2 clusters
Fit 4 clusters
Fit 6 clusters
Fit 8 clusters
Fit 10 clusters
Fit 12 clusters
Fit 14 clusters
Fit 16 clusters
Fit 18 clusters
Fit 20 clusters
Fit 2 clusters
Fit 4 clusters
Fit 6 clusters
Fit 8 clusters
Fit 10 clusters
Fit 12 clusters
Fit 14 clusters
Fit 16 clusters
Fit 18 clusters
Fit 20 clusters

# set k to 6
clusters_assad = MiniBatchKMeans(n_clusters=6, init_size=1024, batch_size=2048, random_state=20).fit_predict(text_Assad)
clusters_ceo =MiniBatchKMeans(n_clusters=6, init_size=1024, batch_size=2048, random_state=20).fit_predict(text_ceo)
clusters_capitol =MiniBatchKMeans(n_clusters=6, init_size=1024, batch_size=2048, random_state=20).fit_predict(text_capitol)
clusters_depp =MiniBatchKMeans(n_clusters=6, init_size=1024, batch_size=2048, random_state=20).fit_predict(text_depp)

#probably overrated
def get_top_keywords(data, clusters, labels, n_terms):
    df = pd.DataFrame(data.todense()).groupby(clusters).mean()
    
    for i,r in df.iterrows():
        print('\nCluster {}'.format(i))
        print(','.join([labels[t] for t in np.argsort(r)[-n_terms:]]))
            
get_top_keywords(text_Assad, clusters_assad, tfidf.get_feature_names_out(), 15)
get_top_keywords(text_ceo, clusters_ceo, tfidf.get_feature_names_out(), 15)
get_top_keywords(text_capitol, clusters_capitol, tfidf.get_feature_names_out(), 15)
get_top_keywords(text_depp, clusters_depp, tfidf.get_feature_names_out(), 15)

Cluster 0
fan,fanatical,going,christians,say,need,live,jews,rid,earth,born,cancer,peacefully,raised,islam

Cluster 1
intervention,attacks,actions,opinion,sending,exchange,demands,urge,assistance,approved,crown,prince,minorities,iran,israel

Cluster 2
fascism,fascist,fast,fat,fate,fanatics,fairly,like,don,coming,greater,learn,isreal,arabs,mistakes

Cluster 3
new,russia,christians,good,rebels,going,syrian,country,home,just,people,like,israel,assad,syria

Cluster 4
families,family,fan,fanatical,fanatics,far,farts,fascism,fascist,familiar,zone,does,end,times,seen

Cluster 5
taking,future,religious,ethnic,opportunity,sides,idk,optimistic,tolerance,expand,thier,pass,im,known,isnt

Cluster 0
hear,issue,understanding,class,seriously,press,cheap,sucks,truth,probably,shouting,wasn,people,like,screaming

Cluster 1
know,media,murder,left,think,man,rich,right,ceo,don,guy,just,like,ben,people

Cluster 2
regards,college,earlier,right,great,got,sure,family,understand,talking,school,masters,degree,comes,kid

Cluster 3
sorry,dead,ceo,health,justice,shame,police,coverage,company,mr,alive,particular,reported,sympathy,individual

Cluster 4
fanatical,fanatics,familiar,country,don,freedom,love,anymore,shame,people,defense,uncle,way,feel,like

Cluster 5
industry,company,pay,denied,united,profit,free,ceo,people,companies,care,health,luigi,insurance,healthcare

Cluster 0
family,fan,fanatical,fanatics,far,farts,fascism,fascist,false,just,media,president,forget,wil,swamp

Cluster 1
finally,just,million,simple,work,visit,honestly,amazing,think,great,comment,thank,video,thought,second

Cluster 2
police,american,video,don,country,right,biden,ben,election,just,supporters,like,america,people,trump

Cluster 3
like,ben,law,feel,political,shapiro,reality,bases,fast,totally,facts,views,holds,beliefs,ideals

Cluster 4
families,family,fan,fanatical,fanatics,far,farts,fascism,help,needs,lies,makes,stuff,attention,mental

Cluster 5
look,day,broke,peaceful,mean,know,right,deceived,don,like,people,group,trump,dressed,supporters

Cluster 0
takes,woman,person,didnt,disgusting,public,eye,insane,school,ive,arent,genuinely,just,doing,annoying

Cluster 1
man,abused,people,did,like,just,case,men,evidence,abuse,woman,cnn,women,heard,trial

Cluster 2
fast,fascist,fascism,family,far,farts,fanatics,fanatical,zone,wannabe,married,time,doing,crazy,job

Cluster 3
families,family,fan,fanatical,fanatics,far,farts,fascism,false,zone,analysis,channel,going,foot,disgusting

Cluster 4
family,fan,fanatical,fanatics,far,farts,fascism,fascist,fast,families,american,bs,high,lady,bar

Cluster 5
failure,fanatics,fanatical,far,like,going,time,happy,cnn,supports,media,mainstream,ah,news,fox

capitol_comments['Likes'] = ['0-10' if i < 10 else '11-100' if i < 100 else '101-500' if i < 500 else '501-1000' for i in capitol_comments['Likes']]
depp_comments['Likes'] = ['0-10' if i < 10 else '11-100' if i < 100 else '101-500' if i < 500 else '501-1000' for i in depp_comments['Likes']]
ceo_comments['Likes'] = ['0-10' if i < 10 else '11-100' if i < 100 else '101-500' if i < 500 else '501-1000' for i in ceo_comments['Likes']]
assad_comments['Likes'] = ['0-10' if i < 10 else '11-100' if i < 100 else '101-500' if i < 500 else '501-1000' for i in assad_comments['Likes']]

depp_comments['Likes'].hist()

<Axes: >

ceo_comments['Likes'].hist()

<Axes: >

assad_comments['Likes'].hist()

<Axes: >

capitol_comments['Likes'].hist()

<Axes: >

assad_comments.head()

analyser = SentimentIntensityAnalyzer()

depp_comments.groupby(['leaning', 'Likes', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Johnny Depp Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

capitol_comments.groupby(['leaning', 'Likes', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Capitol Attack Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

ceo_comments.groupby(['leaning', 'Likes', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='CEO Assasination Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

assad_comments.groupby(['leaning', 'Likes', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Assad Fall Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

<Axes: title={'center': 'Assad Fall Topic distribution'}, xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments'>

depp_comments.groupby(['leaning', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Johnny Depp Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

capitol_comments.groupby(['leaning', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Capitol Attack Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

ceo_comments.groupby(['leaning', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='CEO Assasination Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

assad_comments.groupby(['leaning', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Assad Fall Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

<Axes: title={'center': 'Assad Fall Leaning against Channel Topic distribution'}, xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments'>

def calculate_polarity_scores(data: pd.DataFrame):
    compval1 = [ ] # empty list to hold our computed 'compound' VADER scores
    i=0 # counter

    while (i<len(data)):

        k = analyser.polarity_scores(data.iloc[i]['Comment'])
        compval1.append(k['compound'])
        
        i = i+1
        
    # convert sentiment values to numpy for easier usage
    compval1 = np.array(compval1)
    print(len(compval1))
    print(compval1[0])
    data['Polarity_scores'] = compval1

calculate_polarity_scores(assad_comments)
calculate_polarity_scores(ceo_comments)
calculate_polarity_scores(capitol_comments)
calculate_polarity_scores(depp_comments)

10949
-0.8217
11022
0.1779
64683
-0.6983
26612
-0.8858

ceo_comments

# assign score categories and logic
global i

def predict_sentiment(data: pd.DataFrame):
    i = 0
    predicted_value = [ ] # empty series to hold our predicted values

    while(i<len(data)):
        if ((data.iloc[i]['Polarity_scores'] >= 0.7)):
            predicted_value.append('positive')
            i = i+1
        elif ((data.iloc[i]['Polarity_scores'] > 0) & (data.iloc[i]['Polarity_scores'] < 0.7)):
            predicted_value.append('neutral')
            i = i+1
        elif ((data.iloc[i]['Polarity_scores'] <= 0)):
            predicted_value.append('negative')
            i = i+1

    data['predicted sentiment'] = predicted_value

predict_sentiment(assad_comments)
predict_sentiment(ceo_comments)
predict_sentiment(capitol_comments)
predict_sentiment(depp_comments)

ceo_comments

ceo_plot = ceo_comments.groupby('predicted sentiment').size().plot(kind='bar')
ceo_plot.set_title('CEO Sentiment Type')

Text(0.5, 1.0, 'CEO Sentiment Type')

capitol_plot = capitol_comments.groupby('predicted sentiment').size().plot(kind='bar')
capitol_plot.set_title('Capitol Sentiment Type')

Text(0.5, 1.0, 'Capitol Sentiment Type')

assad_plot = assad_comments.groupby('predicted sentiment').size().plot(kind='bar')
assad_plot.set_title('Assad Sentiment Type')

Text(0.5, 1.0, 'Assad Sentiment Type')

depp_plot = depp_comments.groupby('predicted sentiment').size().plot(kind='bar')
depp_plot.set_title('Depp Sentiment Type')

Text(0.5, 1.0, 'Depp Sentiment Type')

from wordcloud import WordCloud,STOPWORDS

depp_comments.groupby(['predicted sentiment', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Johnny Depp Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')
ceo_comments.groupby(['predicted sentiment', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='CEO Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')
capitol_comments.groupby(['predicted sentiment', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Capitol Attack Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')
assad_comments.groupby(['predicted sentiment', 'channel']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Assad Fall Leaning against Channel Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

<Axes: title={'center': 'Assad Fall Leaning against Channel Topic distribution'}, xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments'>

def get_top_keywords(data, sentiment):
    stop_words = set(stopwords.words('english'))
    comments = data[data['predicted sentiment'] == sentiment]['Comment']
    all_words = ' '.join(comments).lower()
    words = word_tokenize(all_words)
    filtered_words = [word for word in words if word.isalnum() and word not in stop_words]
    word_counts = Counter(filtered_words)
    top_keywords = word_counts.most_common(40)
    return top_keywords

channels = ['depp_comments', 'ceo_comments', 'capitol_comments', 'assad_comments']
sentiments = ['negative', 'neutral', 'positive']

for channel in channels:
    data = globals()[channel]
    print(f"Top 40 keywords for {channel}:")
    for sentiment in sentiments:
        keywords = get_top_keywords(data, sentiment)
        print(f"\n{sentiment.capitalize()} comments:")
        for word, count in keywords:
            print(f"{word}: {count}")

Top 40 keywords for depp_comments:

Negative comments:
amber: 5382
women: 3637
heard: 3488
depp: 3031
johnny: 2906
trial: 2798
abuse: 2686
people: 2607
woman: 2345
men: 2302
evidence: 2247
like: 2183
one: 2077
case: 2062
cnn: 1869
get: 1813
even: 1809
would: 1802
victim: 1532
man: 1516
believe: 1512
abused: 1368
think: 1340
know: 1311
victims: 1283
never: 1228
abuser: 1211
see: 1151
got: 1148
time: 1143
watch: 1140
lies: 1101
also: 1083
right: 999
back: 994
really: 989
said: 985
say: 981
jury: 977
lied: 962

Neutral comments:
amber: 1112
like: 1044
depp: 795
johnny: 787
women: 743
trial: 723
people: 717
heard: 716
get: 539
would: 517
woman: 515
cnn: 492
one: 486
men: 483
case: 474
think: 406
evidence: 397
even: 382
man: 355
watch: 332
see: 329
never: 329
truth: 327
know: 322
want: 312
good: 305
time: 298
right: 285
really: 278
video: 275
believe: 273
marriage: 272
lol: 267
got: 266
media: 262
jury: 261
justice: 255
way: 253
said: 251
court: 247

Positive comments:
like: 797
people: 644
amber: 580
johnny: 564
depp: 502
women: 491
love: 447
men: 432
would: 426
heard: 415
marriage: 409
get: 399
one: 388
woman: 360
trial: 359
case: 359
man: 345
think: 314
know: 313
good: 307
truth: 306
even: 267
see: 260
also: 257
win: 255
really: 254
want: 241
never: 239
time: 236
evidence: 233
believe: 222
god: 219
justice: 219
life: 216
make: 210
way: 195
need: 195
much: 190
could: 185
say: 184
Top 40 keywords for ceo_comments:

Negative comments:
people: 1649
insurance: 1075
healthcare: 970
get: 655
health: 638
ceo: 637
like: 636
system: 577
ben: 548
one: 538
us: 494
would: 491
care: 466
companies: 460
right: 450
money: 387
make: 386
murder: 385
guy: 385
man: 374
even: 367
know: 362
luigi: 351
many: 351
think: 348
time: 315
life: 314
media: 308
left: 301
need: 293
pay: 293
denied: 288
never: 279
back: 279
violence: 277
wrong: 273
going: 268
medical: 265
american: 263
way: 258

Neutral comments:
people: 555
healthcare: 437
like: 432
insurance: 389
care: 382
health: 381
luigi: 275
get: 225
free: 220
would: 217
system: 210
us: 209
ben: 197
one: 182
ceo: 177
rich: 160
companies: 160
money: 156
right: 156
united: 141
think: 136
need: 125
guy: 124
time: 123
make: 122
want: 119
know: 118
pay: 118
going: 113
medical: 112
see: 112
many: 111
trump: 110
good: 108
never: 104
left: 102
much: 100
way: 100
go: 97
even: 97

Positive comments:
care: 367
people: 333
health: 311
like: 264
healthcare: 232
insurance: 200
us: 174
rich: 141
get: 138
system: 137
one: 136
good: 134
would: 128
right: 126
free: 119
money: 112
united: 109
profit: 104
think: 104
ben: 98
need: 96
love: 93
even: 90
companies: 90
make: 88
luigi: 86
ceo: 86
know: 84
also: 80
life: 79
government: 79
better: 77
country: 77
pay: 75
time: 73
man: 73
best: 73
left: 73
go: 73
way: 72
Top 40 keywords for capitol_comments:

Negative comments:
trump: 9331
people: 8818
like: 3115
america: 3067
election: 2778
us: 2743
one: 2733
would: 2664
country: 2575
right: 2529
get: 2522
ben: 2428
antifa: 2413
biden: 2213
blm: 2161
capitol: 2035
see: 1963
police: 1919
even: 1910
think: 1903
going: 1781
president: 1779
american: 1772
time: 1754
know: 1719
left: 1706
media: 1696
violence: 1624
say: 1610
supporters: 1602
government: 1578
go: 1571
democracy: 1482
never: 1468
way: 1466
world: 1465
news: 1421
fraud: 1402
need: 1378
video: 1365

Neutral comments:
trump: 3351
like: 2548
people: 2479
america: 1284
would: 1141
supporters: 1004
country: 993
us: 968
ben: 901
one: 853
good: 808
get: 796
video: 795
antifa: 762
right: 754
biden: 747
see: 707
think: 683
world: 657
want: 642
peaceful: 626
american: 621
well: 620
capitol: 614
party: 598
know: 578
election: 577
police: 573
time: 571
blm: 571
lol: 570
even: 563
democracy: 562
president: 560
great: 544
going: 536
left: 523
go: 509
never: 495
really: 489

Positive comments:
trump: 1769
people: 1611
like: 1409
ha: 1343
party: 714
america: 691
would: 684
us: 668
good: 647
one: 615
great: 608
right: 587
supporters: 580
country: 579
video: 514
get: 509
think: 500
love: 495
biden: 461
god: 452
world: 442
election: 439
ben: 432
president: 429
well: 424
need: 419
even: 413
american: 406
see: 399
know: 399
make: 382
left: 368
say: 367
time: 365
really: 357
way: 356
want: 355
going: 341
take: 334
states: 333
Top 40 keywords for assad_comments:

Negative comments:
syria: 1258
assad: 834
people: 687
israel: 622
us: 574
war: 450
back: 429
go: 403
country: 392
syrian: 386
like: 382
going: 349
rebels: 342
russia: 341
one: 323
another: 322
get: 305
world: 289
new: 285
trump: 272
terrorist: 266
regime: 255
middle: 243
christians: 240
time: 239
years: 236
would: 235
see: 231
iran: 228
isis: 222
even: 219
know: 218
east: 217
terrorists: 212
never: 207
west: 202
syrians: 200
bad: 196
home: 193
think: 189

Neutral comments:
syria: 606
like: 505
israel: 310
assad: 306
people: 288
go: 220
back: 203
good: 193
us: 193
syrian: 191
country: 189
one: 144
new: 139
ben: 135
rebels: 131
get: 130
going: 129
christians: 127
another: 124
would: 124
home: 116
know: 114
want: 107
russia: 107
god: 106
better: 105
middle: 104
world: 102
hope: 102
east: 97
take: 95
well: 94
looks: 93
great: 92
never: 92
regime: 91
thank: 89
time: 87
years: 83
isis: 83

Positive comments:
syria: 360
people: 290
like: 219
assad: 166
israel: 155
country: 134
god: 133
free: 131
syrian: 131
good: 126
peace: 117
hope: 111
us: 110
one: 108
go: 93
great: 85
happy: 83
would: 83
freedom: 81
better: 81
world: 80
back: 76
see: 76
president: 75
christians: 72
time: 71
get: 71
years: 71
well: 71
jesus: 71
ben: 71
new: 70
love: 70
let: 69
regime: 67
going: 67
much: 64
also: 61
even: 61
many: 60

import pandas as pd
import random
import nltk
from nltk.sentiment.vader import SentimentIntensityAnalyzer
from collections import Counter
from nltk.corpus import stopwords
from nltk.tokenize import word_tokenize

# Download the VADER lexicon and stopwords
nltk.download('vader_lexicon')
nltk.download('stopwords')
nltk.download('punkt')

# Initialize the VADER sentiment analyzer
analyser = SentimentIntensityAnalyzer()

# Function to get top 50 unique keywords for each sentiment
def get_unique_keywords(data, sentiment):
    stop_words = set(stopwords.words('english'))
    comments = data[data['predicted sentiment'] == sentiment]['Comment']
    all_words = ' '.join(comments).lower()
    words = word_tokenize(all_words)
    filtered_words = [word for word in words if word.isalnum() and word not in stop_words]
    word_counts = Counter(filtered_words)
    top_keywords = word_counts.most_common(50)
    return set([word for word, count in top_keywords])

# Function to get unique keywords for each sentiment
def get_sentiment_keywords(data):
    negative_keywords = get_unique_keywords(data, 'negative')
    neutral_keywords = get_unique_keywords(data, 'neutral')
    positive_keywords = get_unique_keywords(data, 'positive')

    unique_negative = negative_keywords - neutral_keywords - positive_keywords
    unique_neutral = neutral_keywords - negative_keywords - positive_keywords
    unique_positive = positive_keywords - negative_keywords - neutral_keywords

    return unique_negative, unique_neutral, unique_positive

# Assuming your data is in DataFrames called depp_comments, ceo_comments, capitol_comments, and assad_comments
channels = ['depp_comments', 'ceo_comments', 'capitol_comments', 'assad_comments']

for channel in channels:
    data = globals()[channel]
    unique_negative, unique_neutral, unique_positive = get_sentiment_keywords(data)
    print(f"Top 50 unique keywords for {channel}:")
    print(f"\nNegative comments: {list(unique_negative)[:50]}")
    print(f"\nNeutral comments: {list(unique_neutral)[:50]}")
    print(f"\nPositive comments: {list(unique_positive)[:50]}")

[nltk_data] Downloading package vader_lexicon to
[nltk_data]     C:\Users\X\AppData\Roaming\nltk_data...
[nltk_data]   Package vader_lexicon is already up-to-date!
[nltk_data] Downloading package stopwords to
[nltk_data]     C:\Users\X\AppData\Roaming\nltk_data...
[nltk_data]   Package stopwords is already up-to-date!
[nltk_data] Downloading package punkt to
[nltk_data]     C:\Users\X\AppData\Roaming\nltk_data...
[nltk_data]   Package punkt is already up-to-date!

Top 50 unique keywords for depp_comments:

Negative comments: ['go', 'abused', 'going', 'victim', 'lying', 'victims', 'wrong', 'abuse', 'domestic', 'lies', 'lied', 'abuser', 'still']

Neutral comments: ['video', 'lol', 'money', 'thank']

Positive comments: ['life', 'married', 'god', 'great', 'always', 'years', 'need']
Top 50 unique keywords for ceo_comments:

Negative comments: ['company', 'murder', 'take', 'person', 'someone', 'pain', 'back', 'denied', 'wrong', 'violence', 'bad']

Neutral comments: ['americans', 'trump']

Positive comments: ['really', 'government', 'great', 'country', 'best', 'love']
Top 50 unique keywords for capitol_comments:

Negative comments: ['democrats', 'war', 'years', 'many', 'let', 'capital', 'news', 'violence', 'fraud', 'evidence', 'day']

Neutral comments: ['lol', 'peaceful']

Positive comments: ['ha', 'also', 'god', 'republican', 'states', 'better', 'freedom']
Top 50 unique keywords for assad_comments:

Negative comments: ['terrorists', 'dictator', 'islamic', 'war', 'next', 'libya', 'al', 'bad', 'iran', 'terrorist']

Neutral comments: ['refugees', 'looks', 'help', 'turkey', 'want']

Positive comments: ['much', 'also', 'president', 'many', 'let', 'jesus', 'best', 'love', 'peace', 'happy', 'said', 'freedom']

depp_comments.groupby(['predicted sentiment', 'leaning']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Johnny Depp Leaning against sentiment Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')
ceo_comments.groupby(['predicted sentiment', 'leaning']).size().unstack().plot(kind='bar',figsize=(11, 5), title='CEO Leaning against sentiment Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')
capitol_comments.groupby(['predicted sentiment', 'leaning']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Capitol Attack Leaning against sentiment Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')
assad_comments.groupby(['predicted sentiment', 'leaning']).size().unstack().plot(kind='bar',figsize=(11, 5), title='Assad Fall Leaning against sentiment Topic distribution', xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments')

<Axes: title={'center': 'Assad Fall Leaning against sentiment Topic distribution'}, xlabel='Distribution of interest based on media, political spectrum and Likes', ylabel='Number of Comments'>

import sklearn

my_punctuation = '!"$%&\'()*+,-./:;<=>?[\\]^_`{|}~•@'
my_stopwords = nltk.corpus.stopwords.words('english')
word_rooter = nltk.stem.snowball.PorterStemmer(ignore_stopwords=False).stem

def clean_comment(comment, bigrams=False):
    comment = comment.lower() # lower case
    comment = re.sub('['+my_punctuation + ']+', ' ', comment) # strip punctuation
    comment = re.sub('\s+', ' ', comment) #remove double spacing
    comment = re.sub('([0-9]+)', '', comment) # remove numbers
    comment_token_list = [word for word in comment.split(' ')
                            if word not in my_stopwords] # remove stopwords

    comment_token_list = [word_rooter(word) if '#' not in word else word
                        for word in comment_token_list] # apply word rooter
    if bigrams:
        comment_token_list = comment_token_list+[comment_token_list[i]+'_'+ comment_token_list[i+1]
                                            for i in range(len(comment_token_list)-1)]
    comment = ' '.join(comment_token_list)
    return comment

depp_comments['clean_comment'] = depp_comments.Comment.apply(clean_comment)
ceo_comments['clean_comment'] = ceo_comments.Comment.apply(clean_comment)
capitol_comments['clean_comment'] = capitol_comments.Comment.apply(clean_comment)
assad_comments['clean_comment'] = assad_comments.Comment.apply(clean_comment)
assad_comments.head()

from sklearn.feature_extraction.text import CountVectorizer

# the vectorizer object will be used to transform text to vector form
vectorizer = CountVectorizer(max_df=0.9, min_df=25, token_pattern='\w+|\$[\d\.]+|\S+')

# apply transformation
tf_assad = vectorizer.fit_transform(assad_comments['clean_comment']).toarray()

# tf_feature_names tells us what word each column in the matric represents
tf_assad_feature_names = vectorizer.get_feature_names_out()

from sklearn.decomposition import LatentDirichletAllocation

# let us assume 10 topics
number_of_topics = 10

model = LatentDirichletAllocation(n_components=number_of_topics, random_state=0)
# apply the model
model.fit(tf_assad)

LatentDirichletAllocation(random_state=0)

LatentDirichletAllocation(random_state=0)

def display_topics(model, feature_names, no_top_words):
    topic_dict = {}
    for topic_idx, topic in enumerate(model.components_):
        topic_dict["Topic %d words" % (topic_idx)]= ['{}'.format(feature_names[i])
                        for i in topic.argsort()[:-no_top_words - 1:-1]]
        topic_dict["Topic %d weights" % (topic_idx)]= ['{:.1f}'.format(topic[i])
                        for i in topic.argsort()[:-no_top_words - 1:-1]]
    return pd.DataFrame(topic_dict)

no_top_words = 10

# apply transformation
tf_depp = vectorizer.fit_transform(depp_comments['clean_comment']).toarray()

# tf_feature_names tells us what word each column in the matric represents
tf_depp_feature_names = vectorizer.get_feature_names_out()

model = LatentDirichletAllocation(n_components=number_of_topics, random_state=0)
# apply the model
model.fit(tf_depp)

no_top_words = 10
display_topics(model, tf_depp_feature_names, no_top_words)

# apply transformation
tf_ceo = vectorizer.fit_transform(ceo_comments['clean_comment']).toarray()

# tf_feature_names tells us what word each column in the matric represents
tf_ceo_feature_names = vectorizer.get_feature_names_out()

model = LatentDirichletAllocation(n_components=number_of_topics, random_state=0)
# apply the model
model.fit(tf_ceo)

no_top_words = 10
display_topics(model, tf_ceo_feature_names, no_top_words)

# apply transformation
tf_capitol = vectorizer.fit_transform(capitol_comments['clean_comment']).toarray()

# tf_feature_names tells us what word each column in the matric represents
tf_capitol_feature_names = vectorizer.get_feature_names_out()

model = LatentDirichletAllocation(n_components=number_of_topics, random_state=0)
# apply the model
model.fit(tf_capitol)

no_top_words = 10
display_topics(model, tf_capitol_feature_names, no_top_words)

def generate_cloudwords(data: pd.DataFrame, sentiment: str):
    ceo_negative = data[data['predicted sentiment']==f'{sentiment}']

    words = ' '.join(ceo_negative['Comment'])
    cleaned_word = " ".join([word for word in words.split()
                                if 'http' not in word
                                    and not word.startswith('@')
                                    and word != 'RT'
                                ])

    stopwords = set(STOPWORDS)

    wordcloud = WordCloud(stopwords=stopwords,
                        background_color='black',
                        width=3000,
                        height=2500
                        ).generate(cleaned_word)

    plt.figure(1,figsize=(12, 12))
    plt.imshow(wordcloud)
    plt.axis('off')
    plt.title(f'Cloudword for {sentiment} {data}')
    plt.show()

generate_cloudwords(ceo_comments, 'negative')
generate_cloudwords(depp_comments, 'neutral')
generate_cloudwords(assad_comments, 'positive')

	Topic 0 words	Topic 0 weights	Topic 1 words	Topic 1 weights	Topic 2 words	Topic 2 weights	Topic 3 words	Topic 3 weights	Topic 4 words	Topic 4 weights	Topic 5 words	Topic 5 weights	Topic 6 words	Topic 6 weights	Topic 7 words	Topic 7 weights	Topic 8 words	Topic 8 weights	Topic 9 words	Topic 9 weights
0	like	1820.6	men	1369.4	’	3105.6	cnn	2537.6	ah	797.9	abus	5678.5	depp	2436.3	evid	1942.1	watch	946.5	get	1103.4
1	peopl	842.9	women	1342.1	’t	2885.1	news	936.3	jd	670.1	amber	3093.9	johnni	1913.5	trial	1839.9	trial	795.1	go	920.7
2	know	822.4	marriag	1177.7	it	986.1	lie	712.1	women	491.5	victim	2510.3	heard	1278.7	juri	1211.6	video	790.8	money	843.7
3	love	697.8	get	1090.6	i	930.9	watch	700.9	elain	477.5	heard	2125.3	men	1128.5	case	1129.2	peopl	429.6	hit	686.6
4	shit	651.7	peopl	782.5	don	753.0	media	658.2	accus	320.9	lie	1892.8	win	1057.3	watch	1100.8	lol	429.4	pay	683.7
5	amber	633.7	man	781.8	”	736.3	amber	649.0	fals	290.3	women	1857.0	amber	1017.8	amber	1020.1	like	414.4	johnni	659.1
6	realli	631.8	marri	767.8	…	725.6	truth	439.7	lie	279.3	believ	1560.7	women	957.3	lie	939.1	talk	401.8	want	627.9
7	thing	553.9	woman	721.4	’m	594.1	trial	436.3	amber	252.6	johnni	1147.4	case	822.5	uk	913.9	thank	352.2	would	616.5
8	guy	523.8	want	652.3	didn	593.1	liar	388.7	need	210.7	depp	1022.2	abus	804.7	judg	755.5	care	329.6	time	563.2
9	think	482.6	abus	597.0	she	563.0	fake	313.6	woman	204.7	evid	953.4	victim	682.8	bbc	709.3	comment	303.4	got	559.7

	Topic 0 words	Topic 0 weights	Topic 1 words	Topic 1 weights	Topic 2 words	Topic 2 weights	Topic 3 words	Topic 3 weights	Topic 4 words	Topic 4 weights	Topic 5 words	Topic 5 weights	Topic 6 words	Topic 6 weights	Topic 7 words	Topic 7 weights	Topic 8 words	Topic 8 weights	Topic 9 words	Topic 9 weights
0	healthcar	751.8	’t	406.5	insur	1537.4	’	1162.7	ceo	719.5	trump	347.4	ben	659.5	luigi	731.7	care	355.4	like	765.7
1	system	576.8	i	310.1	compani	885.1	’t	687.6	murder	548.8	xd	280.5	right	587.5	free	425.4	american	257.7	look	404.6
2	peopl	537.8	…	227.5	health	884.9	it	432.8	peopl	492.5	guy	164.0	left	484.0	mangion	229.2	us	236.7	guy	294.7
3	countri	288.2	’m	218.1	care	722.7	don	294.3	kill	388.3	peopl	159.8	peopl	261.4	back	205.8	health	229.8	say	234.7
4	us	251.3	guy	153.0	healthcar	636.9	peopl	175.5	man	310.4	stop	133.7	shapiro	204.2	go	188.5	work	210.8	show	229.7
5	chang	219.4	’r	152.8	deni	555.3	he	163.5	famili	284.9	think	118.6	issu	168.8	pain	141.5	peopl	205.6	one	227.8
6	corrupt	214.3	know	126.1	pay	534.8	that	161.5	mani	219.9	tri	117.6	us	152.4	get	135.5	make	188.9	thing	197.5
7	govern	209.0	don	124.7	medic	461.2	insur	141.8	one	219.3	u	116.8	rich	151.3	mental	113.5	money	177.5	know	196.3
8	america	193.1	”	111.6	peopl	450.6	care	132.9	get	202.2	want	114.8	class	150.3	peopl	94.4	system	171.1	make	184.2
9	american	190.5	😂	111.4	get	375.8	compani	121.0	death	190.8	report	105.8	media	148.3	black	84.4	healthcar	132.1	would	173.6

	Topic 0 words	Topic 0 weights	Topic 1 words	Topic 1 weights	Topic 2 words	Topic 2 weights	Topic 3 words	Topic 3 weights	Topic 4 words	Topic 4 weights	Topic 5 words	Topic 5 weights	Topic 6 words	Topic 6 weights	Topic 7 words	Topic 7 weights	Topic 8 words	Topic 8 weights	Topic 9 words	Topic 9 weights
0	elect	3398.3	protest	2858.8	video	2425.4	trump	6286.5	peopl	3626.2	’	3445.7	america	3119.9	news	1673.0	evid	1597.3	day	1591.7
1	vote	2985.2	peopl	1954.9	right	1525.1	support	4135.7	media	1130.7	’t	3296.0	trump	2855.0	war	1194.3	trump	1399.0	ha	1407.1
2	peopl	2442.7	polic	1865.0	parti	1453.5	antifa	3591.6	countri	1090.1	trump	2310.0	world	2208.7	civil	883.7	elect	1342.5	countri	1402.8
3	biden	2241.0	white	1625.6	like	1342.9	riot	2228.2	right	1072.2	ben	2069.3	us	1841.4	fox	869.7	fraud	1196.9	like	1261.6
4	democrat	1435.6	black	1326.3	xd	988.1	blm	2098.1	american	1037.2	peopl	1769.0	countri	1783.2	year	850.6	state	1102.5	watch	1118.1
5	trump	1407.6	peac	1252.0	flag	933.1	capitol	1876.4	take	959.2	it	1361.1	presid	1563.6	start	791.4	court	951.9	democraci	1098.3
6	republican	1180.0	terrorist	1219.0	one	801.0	build	1501.2	left	958.7	like	1310.5	american	1424.2	go	741.2	coup	947.5	america	1044.3
7	would	1114.0	call	1128.2	good	781.5	peopl	1296.8	govern	957.7	i	1291.1	god	1315.4	see	708.8	ben	705.7	american	863.2
8	presid	1113.7	would	1047.9	thought	730.9	burn	1178.5	back	915.7	don	1178.1	usa	1275.2	fake	658.7	lie	676.4	worst	814.5
9	go	1073.8	blm	1011.8	left	712.6	protest	1057.0	work	872.2	say	1100.5	state	1263.5	get	652.3	bbc	565.6	lol	700.3

Youtube Comments Sentiment Analysis

Cleaning dataset

Split dataset into multiple subsets based on the topic

Text Clustering

Sentiment Analysis

Number of Comments by size performed by overall group and separated channels

NLP Polarization Topic Modeling

Distribution of Sentiment Type by CEO Assassination Topic

Distribution of Sentiment Type by Capitol Attack Topic

Distribution of Sentiment Type by Assad Fall Topic

Distribution of Sentiment Type by Johnny Depp Trial Topic

Distribution of Sentiment by channel and sentiment leaning

Topic Modeling by LDA algorithm

Topics weights based on multiple words

Cloud Words for most mentioned topics

	User	Comment	Published At	Time Ago	channel	topic	leaning
0	@joseaugustofigueiredo2796	BBC journalists must go to the doors of Britis...	2024-12-17 11:29:42	2024-12-17 11:29:42	bbc	assad	news
1	@AzadBus-h1h	Getting out of hands. It may sound off topic b...	2024-12-15 21:38:09	2024-12-15 21:38:09	bbc	assad	news
2	@Emmanuel-n7y	Imperialist go home	2024-12-14 08:39:23	2024-12-14 08:39:23	bbc	assad	news
3	@mrb6309	It's like we've seen this movie in the Middle ...	2024-12-13 05:23:05	2024-12-13 05:23:05	bbc	assad	news
4	@musadube1647	😂😂😂 how dumb and foolish the Americans and the...	2024-12-13 01:29:48	2024-12-13 01:29:48	bbc	assad	news

	Comment	Likes	channel	leaning
0	bbc journalists must go to the doors of britis...	0-10	bbc	news
1	getting out of hands. it may sound off topic b...	0-10	bbc	news
2	imperialist go home	0-10	bbc	news
3	it's like we've seen this movie in the middle ...	0-10	bbc	news
4	😂😂😂 how dumb and foolish the americans and the...	0-10	bbc	news

	Comment	Likes	channel	leaning	Polarity_scores
3237	well he was stressed out so of course he went ...	0-10	cnn	news	0.1779
3238	police were being forceful with him . luig...	0-10	cnn	news	-0.4545
3239	nullify nullify \nfree luigi	0-10	cnn	news	0.5106
3240	i am sad!! wish him well!!	0-10	cnn	news	0.4344
3241	hope we can save luigi.	0-10	cnn	news	0.7269
...	...	...	...	...	...
108094	his hoodie was a mid tone hue that's why it lo...	0-10	young_turks	centre	0.0000
108095	_“those who make peaceful revolution impossibl...	11-100	young_turks	centre	-0.5106
108096	so young, so handsome, and an ivy league compu...	0-10	young_turks	centre	-0.1416
108097	i'm tellin' y'all, that's not luigi! that's su...	11-100	young_turks	centre	0.0000
108098	the nypd said this guy is black😂	0-10	young_turks	centre	0.0000